Do km-scale models better simulate near-surface winds?

Realistic simulations of near-surface wind speeds are important for many reasons, including an accurate characterisation of storm effects on dust-particle emissions. Km-scale models are expected to represent winds including their extremes more realistically by explicitly resolving mesoscale dynamics; however, the extent to which they outperform coarser-resolution models has not yet been systematically assessed. In this study, we conduct a multi-dataset, multi-resolution comparison of sub-daily near-surface wind speeds and the dust uplift potential (DUP) for North African dust regions for the period 1994–2014. The analysis integrates recently developed global km-scale climate simulations from ICON (Icosahedral Nonhydrostatic) and IFS (Integrated Forecasting System), reanalysis products including ERA5 (ECMWF Reanalysis v5) and MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, Version 2), historical climate simulations from CMIP5 and CMIP6 (Coupled Model Intercomparison Projects), as well as observational data from surface meteorological stations. In addition to statistical analyses of the sub-daily winds across these datasets, we have applied a machine-learning technique to pinpoint the weather patterns that drive wind differences across the models.

The results highlight that the two kilometre-scale models ICON and IFS show an overall improved representation of observed surface wind speed distributions, along with reanalysis products, compared to coarser-resolution CMIP models. However, the level of agreement varies with region, season, and time of day. For instance, winds in the Sahel region show higher consistency with observed wind speed distributions for all models, whereas substantially larger deviations occur over the Bodélé Depression, which is the world’s most active dust source, in the coarser-resolution simulations of CMIP compared to observations. The largest inter-model differences are seen during boreal winter (December–February), when northeasterly Harmattan winds often occur, and are most pronounced during the early morning hours (06 - 09 UTC), pointing to the breakdown of nocturnal low-level jets. This work provides an assessment of the strengths and limitations of contemporary global datasets for simulating dust-relevant winds over North Africa and provides a reference framework for evaluating upcoming model output from CMIP7 historical experiments.